AuthMem-Bench (Unauthorized Action Rate): leaderboard
Metric: Attack success rate (%; share of 350 non-authorizing cases in which the action model issues the prohibited tool call, matched target tool and full argument object; washed focal memory with no authority metadata). Source: arxiv.org. Saturation forecast: Around 2030. 7 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Pro | 38.6 |
| 2 | Qwen 3.7 Max | 46.6 |
| 3 | Gemini 3.5 Flash | 52 |
| 4 | GLM-5.2 | 52 |
| 5 | Gemini 3.1 Pro (Preview) | 52.6 |
| 6 | GPT-5.5 | 54.3 |
| 7 | GPT-5.4 Mini | 56.3 |
Interactive version: theaggregate.ai/benchmark?slug=authmem-bench-unauthorized-action-rate · How It Works · Data refreshed daily, snapshot 2026-09-26.